Author: Robert Hyatt
Date: 11:45:18 07/15/04
Go up one level in this thread
On July 15, 2004 at 13:29:25, Uri Blass wrote: >On July 15, 2004 at 13:00:07, Eugene Nalimov wrote: > >>On July 15, 2004 at 12:24:37, Uri Blass wrote: >> >>>On July 15, 2004 at 12:16:53, Eugene Nalimov wrote: >>> >>>>On July 15, 2004 at 05:11:21, Uri Blass wrote: >>>> >>>>>On July 14, 2004 at 22:05:25, Robert Hyatt wrote: >>>>> >>>>>>On July 14, 2004 at 14:40:17, Omid David Tabibi wrote: >>>>>> >>>>>>>On July 14, 2004 at 14:17:33, Matthew Hull wrote: >>>>>>> >>>>>>>>On July 14, 2004 at 13:00:15, Omid David Tabibi wrote: >>>>>>>> >>>>>>>>> >>>>>>>>>>The goal is the best computer chess. You can't have that unless it's open >>>>>>>>>>hardware. It has always been open hardware, >>>>>>>>> >>>>>>>>>You forget WMCCC. >>>>>>>> >>>>>>>> >>>>>>>>No I don't. This is not WMCCC. It's WCCC. The best chess will be played on >>>>>>>>big hardware. That's why it's open hardware. >>>>>>>> >>>>>>>>If you want to argue for inferior chess, then go organize thw World Inferior >>>>>>>>Computer Chess Championship. >>>>>>>> >>>>>>>>Meanwhile, the rest of us want to see the best computer chess the world has to >>>>>>>>offer. We want to see the envelope pushed as far as it can go. >>>>>>>> >>>>>>>> >>>>>>>>The best computer chess in the world is supposed to be seen at the World >>>>>>>>Championship. You can't do that by limiting hardware. >>>>>>>> >>>>>>>>I don't know how many different ways it needs to be said. Your idea is fine >>>>>>>>for some other event that is not the WCCC. >>>>>>> >>>>>>>I did not suggest to abolish the open hardware format to begin with. What I >>>>>>>suggest is to hold two events, WCCC for open hardware, and WMCCC for uniform >>>>>>>hardware. Just the way it used to be. In WCCC you will find the best >>>>>>>engine+hardware combination, and in WMCCC you will find the strongest chess >>>>>>>program. >>>>>>> >>>>>> >>>>>> >>>>>>Absolutely and totally bogus statement. >>>>>> >>>>>>What processor will you pick? I want 64 bits. Others want 32 bits. >>>>> >>>>> >>>>>The best solution is to give the participants the possibility to choose the >>>>>hardware when the participants do not need to care to bring the hardware that >>>>>they choose and the organizers do it for them. >>>>> >>>>>If we talk about WMCCC then a better alternative than uniform hardware is that >>>>>programmers will be able to ask the organizers to give them every machine that >>>>>they ask(behind some price) with only one codition that the machine does not >>>>>have more than one cpu. >>>> >>>>Programmer A spent several months to implement parallel search in his program. >>>> >>>>Programmer B spent those months to rewrite his uni-proc search into assembly >>>>language. >>>> >>>>You suggested rule favors programmer B. Why? >>> >>>A can participates in WCCC when B participates in WMCCC >>> >>>I did not say not to have WCCC >> >>Programmer A asked (and got) from the organizers acess to 1-CPU Power5 system >>that costs $20k. > >I think that WMCCC should not be for expensive hardware so I think A should not >get it >and there should be some limitation about the price of the computer that the >organizers give. Uri... I know and understand where you are coming from. But what Eugene (and I and others) are trying to point out is that this has _all_ happened in the past, already. IE the same suggestions. The same questions. The same issues. For example, how do you define the "price" of a given machine? Retail price? Who pays that? Wholesale price? Who defines that. Price to build? What vendor will reveal that? This is all very sticky to deal with. There are far more problems than there are solutions... > >> >>Programmer B asked dual Opteron system that costs less than $3k. Should he get >>it? Chances that average chess program user will get dual Opteron system are >>much higher than chances that she will get Power5 system. > >I suggest that B should get it only for WCCC and not for WMCCC. Based on what? IE by next year, a dual-core single chip opteron will have two cpus even though it has exactly one chip on the MB. Is that not allowed? If not, year after next there will be _no_ computers that will be allowed as it is almost a certainty that every CPU will be dual-core in 2006. So. Back to the question. Allow duals or not? >> >>Programmer C asked for new Pentium 5 (or Pentium 6, or ...) system that have >>*efficient* implementation of multithreading. I.e *one* CPU can run 2 threads >>simultaneously, with effective speedup (say) 1.8x. Should he get it? Cost of >>system is $2.5k. > >I think no. > >All the idea of one cpu for WMCCC is to not to test parallel code that is tested >in WCCC. Then the WMCCC is dead, which is _exactly_ where it is now. And this is _exactly_ the reason it died. The boundary between big boxes and the PC became impossible to pinpoint. You can buy a good dual for no more than a top PIV box easily. It is physically no bigger, it is cheaper, it is highly available as there are hundreds of dual-cpu boards around, etc... > >> >>Programmer D asked for new AMD K9 (or K10, or ...) system that have dual cores >>*on chip*. I.e *one* physical chip contains 2 CPU cores, with effective speedup >>(say) 1.95x. Should he get it? Cost of system is $1.8k. > >Again no. Again, WMCCC is dead because after next year you won't be able to buy a single-core CPU unless you choose to play on a cell phone or PDA. A few years out they will be dual core. > >> >>Programmer E asked for new Itanium4 system that have 8 cores *on chip*. >>Effective speedup is (say) 7.5x. Should he get it? Cost of system is $20k, >>exactly as cost of Power5 system we gave to programmer A. > >Again too expensive so no. > >> >>Programmer F asked for off-the-shelf CPU with built-in FPGA (right now there are >>such CPUs for embedded systems). He can use FPGA to dramatically speed his >>search. Cost of system with such CPU is $300 (CPU/memory/serial port, all the >>interface is running on plain PC). Can he use it? > >Allowed for WCCC but if this is not machine that other can buy(if the FPGA is >some special hardware that the programmer designed) then the answer is no. But if it is special-purpose hardware included in the normal system, it is ok? Even though most programmers will not be using it because it requires special assembly language programming tricks to access it??? >> >>Programmer G brought with him FPGA or ASIC system designed by him. Total cost of >>all components is $1.5k. Can he use it? BTW, his hardware can do 16 independent >>searches in parallel, but there is exactly one CPU in his system. > >No >Again WMCCC is only software competition and you should use some standard >hardware that people can buy. >> >>Programmer H asked some time on $100k S/3090. His program can use only one CPU, >>and single-CPU performance of S/3090 is 2x less than Opteron 2GHz. To partially >>compensate for this his program is written in the S/3090 assembly language, so >>it cannot be ported anywhere. He points that he don't get *any* hardware >>advantage from running on sych a system compared to *any* WMCCC participant, so >>he should be entered into WMCCC. > >Again no > >WMCCC is not about expensive solutions. >> >>Programmer J wrote his program for the $40 "Gameboy Portable" with z80 CPU. >>There are good chances that when he will not get 1st place he would say >>"competition was not fair -- they have hardware advantage". Should he be >>allowed? > >Yes and congratulation for J for winning the tournament because J was the only >participant after the other programmers that you suggested were not allowed to >enter with the hardware that they wanted so they decided not to play:-( > >Uri Never seen anyone decide to "not play" at a WCCC or CCT event because of hardware. It might happen. But in the 1970's I competed knowing I had no hardware to compare to the CDC Cyber 176 used by chess 4.x.. yet I came every year and played...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.